Combinatorial Materials Development - American Chemical Society

sputtering in an off-axis geometry using 2 inch metallic targets in planar magnetron ... can be defined as eie0 EB R = FOM, where ε0 is the permittiv...
1 downloads 0 Views 2MB Size
Downloaded via TUFTS UNIV on July 11, 2018 at 14:43:55 (UTC). See https://pubs.acs.org/sharingguidelines for options on how to legitimately share published articles.

Chapter 3

The Compositional Spread Approach to High¬Dielectric Constant Materials and Materials for Integrated Optics L . F. Schneemeyer*, R. B. van Dover*, C. K. Madsen, and C. L . Claypool Bell Laboratories, Lucent Technologies, 600 Mountain Avenue, M u r r a y Hill, NJ 07974-0636

Introduction High throughput synthetic approaches combined with rapid screening techniques to evaluate materials performance so called combinatorial approaches - offer important advantages for materials investigations. Optical communications and the on-going electronic revolution are driving interest in new inorganic solid state materials (1). With ever shortening product life cycles, effective research methodologies are also of great interest in industrial research. Efficient strategies for materials investigation can greatly improve the odds of a successful materials investigation., particularly when the sample materials are prepared using methods similar to those that would be employed in the actual application. _

Combinatorial approaches to the synthesis of large collections, or "libraries", of molecules with possible biological activity have become a standard tool in the drug discovery and optimization process. Together with rapid evaluation of biological activity using various well-developed automated screening approaches, combinatorial chemistry has had significant impact on searches for molecular entities that match particular profiles of desired structure-property relationships. Because similar incentives exist compelling researchers to identify inorganic materials for specific applications with greater speed, analogous high-throughput synthetic techniques together with efficient screening approaches are of increasing importance in solid state materials investigations. The traditional approach to new materials investigations is typically extremely time consuming

© 2002 American Chemical Society Malhotra; Combinatorial Materials Development ACS Symposium Series; American Chemical Society: Washington, DC, 2002.

49

50 and labor intensive. A materials chemist requires about a day to synthesize a single sample and characterization of that sample may take many more days. Clearly, increasing the number of materials that are studied should improve the odds of a breakthrough materials discovery while certainly leading to an increased understanding of composition/property relationships.

Background The concept of the preparation of large numbers of samples simultaneously has been around since at least the late 1960's. Sawatzky and Kay at I B M used cosputtering to prepare gadolinium iron garnet films in which defect concentrations varied along the length of the film to correlate with structural and magnetic properties (2). This efficient synthetic approach eliminated the time and expense of preparing multiple targets which would have been needed had samples in the series been grown one at a time. In addition, any ambiguities which might have arisen from run-to-run variations were eliminated. However, sample evaluation was still carried out manually although the optical characterization relevant to this study was not particularly time consuming. A broader view to the potential power of a parallel synthesis scheme was proposed by Hanak of R C A Labs in 1970 (3). Hanak carried out reactive sputtering from a sectored target as illustrated in Figure 1. A broad range of experimental systems, including superconductors and magnetic materials were studied by Hanak using this approach (4, 5). However, computer control and automated measurements were not yet available at that time, largely because computers were still very expensive and not widely available. The importance of combinatorial chemistry in the pharmaceutical industry motivated Xiang and Schultz to revisit high-throughput synthesis and screening approaches as applied to materials investigations (6). Studies of high temperature cuprate superconductors, colossal magnetoresistance (CMR) manganites and investigations of new phospors demonstrated the power of the approach. In this discrete combinatorial synthesis, DCS, approach, arrays of samples were prepared by multilayer thin film depositions through masks followed by a high temperature annealing step to achieve mixing and produce homogeneous samples. This high temperature anneal often results in the crystallization of samples. Automated screening allows rapid evaluation of materials properties.

Malhotra; Combinatorial Materials Development ACS Symposium Series; American Chemical Society: Washington, DC, 2002.

51

Figure 1. Sputter target layout used for ternary phase spread deposition in the Hanak approach.

While a combinatorial-type approach to a materials study can quickly produce large numbers of samples and a large quantity of data, successful application of such an approach requires the thoughtful application of traditional scientific principles as outlined in Figure 2. A well-defined problem with an explicit goal, a feasible and appropriate synthetic strategy and a measurement and evaluation protocol are elements of a proper materials investigation. Understanding the issues inherent in a particular problem is vital. Samples should preferably be prepared in a form that is meaningfully related to the form in which they will ultimately be used. Samples should be screened by measurements related to the property of interest. Ideally, the results of measurements can be reduced to a scalar figure of merit (FOM) to allow the visual discerning of trends. Also, the time required to evaluate samples should be comparable to the time required by the synthesis step. Even though large numbers of samples can be studied in a combinatorial-type approach, the large number of possible combinations of elements can quickly overwhelm an investigation even before various possible processing parameters are considered. Thus, the extent of a study must be rationally constrained. For

Malhotra; Combinatorial Materials Development ACS Symposium Series; American Chemical Society: Washington, DC, 2002.

52 example, there are 720 different combinations of 10 elements taken 3 at a time. Constraints must be applied to confine a study within manageable limits.

Figure 2. Criteria for the successful use of high-throughput synthesis/screening.

The Continuous Compositional Spread (CCS) Approach While the D C S approach produces samples in thin-film form, the high temperature annealing step eliminates this approach for studies of low temperature metastable or amorphous materials. We have developed the continuous compositional spread (CCS) approach (7) for investigations of materials deposited at relatively low temperatures. In this section, the C C S approach is described and its use in the discovery of a new thin film high dielectric constant material with high breakdown fields and low leakage currents is discussed. In the C C S approach, we use off-axis cosputtering to produce binary or ternary composition spreads as illustrated in Figure 3. In off-axis deposition onto a fixed substrate, the thickness of the deposited film decreases approximately exponentially with distance from the gun. Test runs of Si-0 and Ta-0 deposition demonstrated that high quality oxide thin films could be obtained by reactive sputtering in an off-axis geometry using 2 inch metallic targets in planar magnetron sputter guns (U.S. Gun II, US, Inc., Campbell, C A ) . A crucial factor, however, was the application of an rf field to the growing film during deposition. This bias causes ion bombardment of the growing film, probably

Malhotra; Combinatorial Materials Development ACS Symposium Series; American Chemical Society: Washington, DC, 2002.

53

vacuum chamber

Figure 3. Schematic drawing of the off-axis deposition system used for the CCS approach.

enhancing the surface mobilities and producing a denser film. Our system allows us to codeposit up to three metals so that we can deposit a large portion of a pseudoternary oxide phase diagram in a single run. Relatively high oxygen partial pressures, 10-40%, were used to ensure fully oxidized films. Note that these phase spreads have inherent thickness variations of approximately a factor of two, a limitation of this method. To know the composition of the film at any position on the phase spread, the offaxis deposition of S i O was studied and modeled. A s noted earlier, thickness decreases exponentially with distance from the gun. However, the 2 inch planar magnetron sources used for these experiments are not point sources making the model somewhat more complicated. Still, fits good to 2-5% were obtained. Overall sputtering rates as well as the decay length varies with sputtering parameters including power, partial pressure of oxygen and total pressure (typically 10 to 50 mtorr). For each material to be sputtered, thickness calibrations measured using profilimetry were carried out for several sets of conditions. Compositions obtained using these models were verified by Rutherford Backscattering (RBS) measurements at several points on the sample. x

Malhotra; Combinatorial Materials Development ACS Symposium Series; American Chemical Society: Washington, DC, 2002.

54

Table I. Comparison of high dielectric constant materials.

Malhotra; Combinatorial Materials Development ACS Symposium Series; American Chemical Society: Washington, DC, 2002.

55 A variety of substrates, those suitable for any sputter deposition technique, can be used for the CCS approach. We find silicon wafers, Si, which are very flat, clean and relatively inexpensive to be particularly convenient. Blanket coats of buffer layers, metals such as platinum, aluminum or Ti/TiN, etc., can be coated on the wafer prior to the deposition of the phase spread. The value of the C C S approach was demonstrated by our discovery of amorphous, high dielectric constant, high breakdown field, low leakage materials in the Zr-Ti-Sn-0 pseudoternary phase diagram. A s integrated circuits continue to migrate to higher levels of integration, it appears likely that alternative high dielectric constant materials will be needed to replace silicon oxide, a-SiO . A t present, designers are resorting either to extremely thin or highly nonplanar aS i O films in capacitor structures used for dynamic random access memory, D R A M . A u s e f u l figure of merit corresponding physically to the maximum charge per unit area that can be stored on a capacitor made of a given material can be defined as eie E R = F O M , where ε is the permittivity of free space and E R is the breakdown field (8). Table I contains the relevant data for materials of current interest as high dielectric constant replacement for a-SiO . While each of these candidate materials has certain advantages, each also suffers limitations. For example, a-TaO , amorphous tantalum oxide films, can be deposited at low temperatures, < 450° C, and can have high breakdown fields and low leakage currents (9). However, its dielectric constant is only modestly higher that that of a-SiO , making it a short-term improvement at best. x

x

0

B

0

B

x

x

x

We confined our CCS search for high dielectric constant, high F O M thin film materials to oxides deposited at low temperatures, < 400° C, consistent with the constraints of backend processing and therefore amorphous. We also limited ourselves to elements compatible with existing IC fabrication processes favoring those already accepted in Si fabrication facilities. Those favored elements include Si, ΑΙ, Ο, N , T i and W . About 30 different elemental composition spreads were evaluated in our initial survey. O f these, the Zr-Ti-Sn-0 system showed evidence for an extensive high F O M region as well as reasonably high specific capacitance values and low leakage, and thus was selected for more detailed examination. Figure 4 shows

Malhotra; Combinatorial Materials Development ACS Symposium Series; American Chemical Society: Washington, DC, 2002.

Figure 4. Raw capacitance and breakdown voltage data for a-Zr-Ti-Sn-O as measured. Malhotra; Combinatorial Materials Development ACS Symposium Series; American Chemical Society: Washington, DC, 2002.

57

Figure 5. Raw figure of merit data for the system Zr-Ti-Sn-0 deposited with guns in the same configuration as Figure 4.

Figure 6. The F O M data for the system Zr-Ti-Sn-0 plotted on a conventional ternary phase diagram. Also shown is the single phase region for the crystalline ceramic of composition Ζη.χβηχΤίΟ^

Malhotra; Combinatorial Materials Development ACS Symposium Series; American Chemical Society: Washington, DC, 2002.

58 the raw capacitance and breakdown voltage as a function of position on the substrate. The positions of the sputter guns are shown schematically. While our films are not of uniform thickness, thickness varies gradually with position and thus valid qualitative trends can be discerned. A thickness independent F O M can be determined by a point-by-point evaluation of the product CV R/A, the capacitance times the breakdown voltage divided by the area of the capacitor structure measured. This quantity is conveniently equal to E^OEBR, the F O M discussed above. The F O M data for the Zr-Ti-Sn-O systems in the same orientation as Figure 4 is shown in Figure 5 together with the leakage current values measured at a stored charge density of 7 μθ/cm . A distinct region of high F O M and a separate region of low leakage are clearly seen. Figure 6 shows the high F O M region (95 percentile contour) mapped onto a conventional ternary phase diagram. This figure also shows the single phase region for the low-loss dielectric ceramic material, Z r i . S n T i 0 , which is used for filter elements in wireless circuits (10). Film studies involving compositions inspired by only by the single phase ceramic region (11,12) miss both the high F O M region and the low loss regions of the phase diagram. Because our high throughput search approach allowed us to search broadly, we found excellent properties in an unexpected region of the composition space. The composition a-Zr T i S n O was identified as a high specific capacitance, high breakdown field and low leakage thin film material that represents a promising solution to the problem of dielectric for future generations of embedded D R A M . B

2

th

x

2

6

2

x

4

x

The Optics Explosion Demand for Internet access and other broadband services has exploded in the late 1990's leading to a push for higher and higher lightwave transmission capacity. The two approaches to providing that capacity are high transmission rates and also wavelength division multiplexing (WDM). The growth rates in optical communication, which are >50% annually, are creating demands for new components for optical communications (13). A dense wavelength division multiplexing, D W D M , optical fiber transmission system increases transmission capacity by combining multiple wavelengths of light onto a single optical fiber which are then separated at the other end as illustrated schematically in Figure 7. Multiplexing and demultiplexing, mux and demux, is accomplished using a P L C devices called arrayed waveguide grating routers, A W G ' s . For long-haul systems, either an erbium or Raman optical

Malhotra; Combinatorial Materials Development ACS Symposium Series; American Chemical Society: Washington, DC, 2002.

59

Figure 7. Simplified schematic drawing of a W D M system.

amplifier is used to boost the signal intensity during transmission. Each wavelength provides a separate channel, which can carry voice, data or video information. While long distance, so-called long haul, transmission systems were the original focus for W D M , an evolution is envisioned in which systems will move from W D M transport to W D M networks. This evolution will involve first the use of fixed, and then, in the more distant future, reconfigurable add-drop nodes. Also, there will be more wavelengths, higher bit rates and more powerful amplifiers as well as new optical amplifiers operating over various regions of the Allwave™ fiber spectrum extending from about 1200 to 1700 nm. Integrated optical components fabricated using variants of well-established silicon processing technology, known as silicon optical bench technology, are likely to play significant roles in future system developments. O f course, several technologies are competing to provide critical capabilities in areas that include routers, reconfigurable add/drop multiplexers, dynamic gain equalizers, dispersion compensators and optical amplifiers (13). Bulk-type devices such as M E M ' s devices and thin-film filters, fiber-type devices such as fiber couplers and planar-type devices such as silicon optical bench and semiconductor devices can all provide various aspects of the needed functionality. While time-tomarket, cost and performance of devices emerging from these different

Malhotra; Combinatorial Materials Development ACS Symposium Series; American Chemical Society: Washington, DC, 2002.

60 approaches will determine their use in systems applications, planar integration has distinct advantages that include compactness and reduced packaging costs, and the use of processes that have already been developed for use in silicon integrated circuit technology. Since planar devices use thin-films, the C C S approach to materials investigations can be used to explore new materials for use in planar lightwave circuits (PLC's). One important function needed in complex W D M optical transmission systems and D W D M (dense wavelength division multiplexed) systems is optical filters. P L C ' s provide an increasingly important filter technology and have the potential, as noted earlier, for continued integration and increased functionality. A n example of a P L C , an arrayed wavelength grating router (AWG) is shown in Figure 8. The wavelengths that are input on a single fiber are split onto multiple fibers by this device as indicated in the figure. A cross-section of a portion of such a device is also shown. A layer of thermal silica is formed on a (100) silicon wafer. P-doped silica is deposited onto the thermal silicon and patterned into waveguide structures. Finally, a Β,Ρ-doped S i 0 glass is deposited as a capping layer. There is about a 2% difference in index between the glass of the waveguide structures and the glasses used for the surrounding base and cladding layers. The use of higher index materials for the core of the waveguides would allow us to make smaller P L C ' s . Indeed, high index waveguides are required for future filter designs such as ring resonators. 2

Ill

Figure 8. Example of a planar lightwave circuit, an arrayed wavelength router.

Malhotra; Combinatorial Materials Development ACS Symposium Series; American Chemical Society: Washington, DC, 2002.

61

source + detector detector

source

sample Figure 9. Schematic representation of the reflectometry instrument.

The CCS approach to optical materials The materials requirements for P L C ' s are films that are highly uniform in index and thickness, have low loss in the infared, 1200-1700 nm, and are compatible with existing silicon processing. Moreover, other film deposition techniques, in particular C V D , should be possible for any material identified. While these requirements must be met for device fabrication, the constraints are relaxed during materials investigations. In particular, thickness and index uniformity must be adequate, typically better than a percent, but only over the size of the device to be fabricated. We have used the CCS approach to examine the addition of metals such as tantalum and zirconium to silica glass to raise the index of refraction. These metals have a single oxidation state, avoiding mixed valence transitions which are fully allowed optical transitions and thus would result in loss. Samples were measure using reflectometry, a technique useful for obtaining index and thickness information on thick films (> 1 μπι). Reflectometry is a non-contact mapping technique that measures index and thickness and instruments are commercially available. As shown schematically in Figure 9, the instrument measures reflected white light simultaneously at two different angles. The ratio of the intensities of these reflected beams can be modeled to give index and thickness.

Malhotra; Combinatorial Materials Development ACS Symposium Series; American Chemical Society: Washington, DC, 2002.

62 Controlling the tantalum content of the glass controls the index in tantalumdoped silica films. The index change from that of pure silica is indicated in Figure 10. Index as a function of composition for the system S i i . T a O was mapped by reflectometry. x

x

y

The difference between the index of refraction of pure amorphous silica, a-Si0 , and the deposited film from which the waveguide structures will be fabricated is known as delta where Δ = Π β - nsi02/nsi02- In this system, delta is proportional to the Ta-doping level. In the tantalum doped silica films, the optical quality of the glass looks good and no scattering is observed. However, loss values still must be determined. Also, any issues related to birefringence are unknown but may be addressed by examining the effect of the introduction of additional substituents. 2

χ ρ

Figure 10. The index as a function of composition for the system Sii. Ta O . x

x

Malhotra; Combinatorial Materials Development ACS Symposium Series; American Chemical Society: Washington, DC, 2002.

y

63 In principle, a parallel study of the performance of a simple optical test structure could be carried out as depicted in Figure 11. B y directly measuring material performance as a function of composition in one experiment, subjects such as processibility, optical loss or materials birefringence could be examined free of run-to-run variations. Such experiments are underway.

Summary The continuous compositional spread approach to new materials investigations has proved its value by the discovery of a new high specific capacitance, high breakdown field, low loss material in the Zr-Ti-Sn-0 phase system. This material may be used for future generations of embedded D R A M structures in integrated circuits. We have also discussed the applicability of the C C S approach to optical materials problems. The use of the CCS approach in the investigation of new high index thin film glasses including the S i i . T a O system for use in planar optical circuits was discussed. More efficient approaches to materials discovery and optimization are likely to be of increasing importance, particularly in the area of optoelectronics. x

x

Figure 11. Scheme for a combinatorial-type approach to device characterization.

Malhotra; Combinatorial Materials Development ACS Symposium Series; American Chemical Society: Washington, DC, 2002.

y

64

References * 1. 2. 3. 4. 5. 6.

7. 8. 9. 10. 11. 12. 13. 14.

Now with Agere Sytems, Murray H i l l , N J 07974. Dagani, R.; Chem & Engin. News 1999, 77, 51-60. Sawatzky, E; Kay, E; IBM J. Res. Develop. 1969, 13, 696-702. Hanak, J. J.; J. Mater. Sci. 1970, 5, 964-971. Hanak, J. J.; Gittleman, J. I; Physica 1971, 55, 555. Hanak, J. J.; Jap. J. Appl. Phys. Suppl. 2, Part 1,1974, 809. Xiang, X . - D ; Sun, X.; Briceno, G . ; Lou, Y.; Wang, K . - A . ; Chang, H.; Wallace-Freedman, W. G.; Chen, S.-W.; Schultz, P. G.; Science, 1995, 268, 1738-1740. van Dover, R. B . , Schneemeyer, L . F., and Fleming, R. M.; Nature, 1998, 392, 162-164. Gerstenberg, D . In Handbook of Thin Film Technology; Maissel, L . ; Glang, R.; eds.; McGraw-Hill, N Y , 1970, Chapter 19, ρ 19-8. Hashimoto, C.; Oikawa, H.; Homma, N.; IEEE Trans. Electron Devices, 1989, 36, 14. Iddles, D . M.; Bell, A . J.; Moulson, A . J.; J. Mater. Sci. 1992, 27, 63036310. Nakagawara, O.; Toyoda, Y.; Kobayashi, M.; Kayayama, Y.; Tabata, H . ; Kawai, T.; J. Appl. Phys. 1998, 80, 388-392. Ramakrishnan, E . S.; Cornett, K . D.; Shapiro, G . H . , Howng, W.-Y. J. Electrochem. Soc. 1998, 145, 358-363. White, A. E . Optics & Photonics News March, 2000, p. 27-30. Li, Yuan P.; Henry, C. H.; in Optical Fiber Telecommunications IIIB (Academic Press: New York) p.319 (1997).

Malhotra; Combinatorial Materials Development ACS Symposium Series; American Chemical Society: Washington, DC, 2002.